1. Field of the Invention
The present invention relates to semiconductor memory devices, and more particularly, to improvement of a circuit for transmitting a driving signal boosted to a potential level exceeding internal power supply voltage to a selected word line. Particularly, the present invention relates to a structure for stabilizing an output voltage of a boosting circuit used for generating a boosted word line driving signal and a structure of an input buffer for receiving an external signal.
2. Description of the Background Art
FIG. 25 schematically shows a structure of a conventional dynamic random access memory. Referring to FIG. 25, a dynamic random access memory includes a memory cell array MA in which memory cells for storing information are arranged in a matrix of rows and columns, an address buffer AB responsive to externally applied address A0-An for generating an internal address, an X decoder ADX for receiving an internal row address from address buffer AB for generating a word line selecting signal selecting a corresponding row of memory cell array MA, a word line driving circuit WD responsive to a word line selecting signal from X decoder ADX for amplifying this word line selecting signal to transmit the same to a selected row (word line), and an Y decoder ADY for receiving an internal column address from address buffer AB for generating a column selecting signal selecting a corresponding column out of memory cell array MA.
Address buffer AB receives in a time-divisional manner a row address specifying a row and a column address specifying a column out of memory cell array MA for generating an internal row address and an internal column address at respective predetermined timings to provide the same to X decoder ADX and Y decoder ADY.
In order to read out data from a memory cell specified by external address A0-An (a memory cell provided corresponding to a crossing of a selected row and column), the dynamic random access memory includes a sense amplifier for detecting and amplifying data of a memory cell connected to a row selected by a word line selecting signal from X decoder ADX and to which a driving signal is transmitted by word line driving circuit WD, and an input/output interface (IO) responsive to a column selecting signal from Y decoder ADY for transmitting to an output buffer OB data of a memory cell connected to a corresponding column out of the memory cells of a selected row (word line). In FIG. 25, the sense amplifier and the input/output interface (IO) are indicated by one block SI.
Output buffer OB generates an external read out data from internal data transmitted via input/output interface (IO) to provide the same external to the device.
Although only output buffer OB for reading out the data is indicated in FIG. 25, an input buffer is also provided for writing data. This input buffer may be implemented to input data from an external device via a pin terminal identical to that of output buffer OB or via a different pin terminal. An input buffer generates an internal write data from an externally applied write data to write the data into a selected memory cell via an input/output interface (I/O).
Control signal generating peripheral circuitry CG for generating control signals to control various operation timings of a dynamic random access memory is provided. Control signal generating peripheral circuitry CG responds to an externally applied control clock signal, i.e. a row address strobe signal /RAS, a column address strobe signal /CAS, and a write enable signal /WE to generate a word line driving signal .phi.x, an equalize signal .phi.E, a precharge signal .phi.p, and sense amplifier activation signals .phi.A and .phi.B. Control signal generating peripheral circuitry CG also generates a precharge potential VB for precharging a bit line or the like to a predetermined potential.
FIG. 26 schematically shows a structure of the memory cell array of FIG. 25 and relating circuitry. Referring to FIG. 26, a memory cell array MA includes a plurality of memory cells 1 arranged in a matrix of rows and columns (n rows, m columns), word lines WL1, WL2, . . . , WLn provided corresponding to the rows of memory cell array MA, and bit lines BL0, /BL0, BL1, /BL1, . . . , BLm, /BLm, provided corresponding to each column in memory cell array MA. A bit line BL (indicating generically bit lines BL0-BLm) and a bit line /BL (indicating generically complementary bit lines /BL0-/BLm) form a folded type complementary bit line pair. One pair of bit lines connects memory cells 1 of one column in memory cell array MA.
In FIG. 26, a bit line BL0 and a complementary bit line /BL0 form a pair of bit lines, and a bit line BL1 and a complementary bit line /BL1 form another pair of bit lines. Similarly, a bit line BLm and a complementary bit line BLm form a pair of bit lines.
A memory cell 1 is provided corresponding to the crossing of one word line and one of the pair of bit lines. More specifically, memory cell 1 is provided corresponding to the crossing of one word line WL (indicating generically word lines WL1-WLn) and one of the bit line pair BL or /BL.
A precharge/equalize (P/E) circuit 150 for equalizing the potential of each bit line at the time of standby of a dynamic random access memory and for precharging to a predetermined potential VB is provided to each pair of bit lines BL0, /BL0, . . . , BLm, /BLm. Each precharge/equalize circuit 150 attains a conducting state in response to a precharge designating signal .phi.p and an equalize designating signal .phi.E to precharge and equalize the potential of each bit lines BL0, /BL0-BLm, /BLm to a predetermined precharge potential VB.
A sense amplifier circuit 160 for detecting and amplifying data of a selected memory cell is provided for each bit line pairs BL and /BL. Sense amplifier circuit 160 is activated in response to a first sense amplifier driving signal .phi.A and a second sense amplifier driving signal .phi.B transmitted via signal lines 162 and 164, respectively, to detect the potential difference of a corresponding bit line pair for differential amplification.
Column selecting gates T0a, T0b, T1a, T1b, . . . Tma, Tmb are provided for each of bit line pairs B10 /BL0, . . . , BLm, /BLm, responsive to column selecting signals Y0-Ym from Y decoder ADY to attain an ON state for connecting a corresponding pair of bit lines to internal data bus DB, /DB. Internal data bus DB and /DB are connected to output buffer OB shown in FIG. 25.
Column selecting gates T0a, T0b are provided corresponding to the pair of bit lines BL0, /BL0. Column selecting gates T1a and T1b are provided corresponding to the pair of bit lines BL1, /BL1. Column selecting gates Tma and Tmb are provided corresponding to the pair of bit lines BLm, /BLm.
Only one of column selecting signals Y0-Ym from Y decoder ADY is activated according to a column address, whereby the corresponding column selecting gate is turned on. As a result, a corresponding pair of bit lines is connected to internal data bus DB and /DB.
FIG. 27 shows the structure of components associated with one word line of FIG. 26, and particularly indicates the specific structure of the circuit driving a word line.
Referring to FIG. 27, memory cell 1 disposed at the crossing of a word line 3 (WLi) and a bit line 2 (BLj) includes a memory capacitor 6 for storing information in the form of charge, and a selecting transistor 5 turned on in response to a word line driving signal .phi.xi transmitted on word line 3 to connect memory capacitor 6 to bit line 2. Selecting transistor 5 is formed of an n channel insulating gate type field effect transistor (referred to simply as "n-FET" hereinafter), having the gate connected to word line 3, the source to bit line 2, and the drain to storage node 4.
Memory capacitor 6 has one electrode connected to the drain of selecting transistor 5 via storage node 4, and the other electrode connected to receive a potential of 1/2 the operating power supply potential Vcc.
Word line 3 is accompanied with a parasitic capacitance 7. Parasitic capacitance 7 also includes the gate capacitance of selecting transistor 5 of memory cell 1.
Corresponding to word line 3 (WLi), a (unit) X decoder ADXi for decoding an internal row address from the address buffer for generating a word line selecting signal for word line 3 (WLi), and a (unit) word line driver WDi for receiving an output of X decoder ADXi via a node 9 and transmitting a word line driving signal .phi.xi on word line 3 are provided.
X decoder ADXi generates a signal of "H" (logical high) on node 9 when selected.
Word line driver WDi includes an n-FET 14 for passing along a signal from X decoder ADXi provided on node 9, an n-FET 11 responsive to a signal on node 15 transmitted from n-FET 14 for transmitting a word line driving signal .phi.x from node 10 to word line 3 via a node 13, an inverter circuit 16 for inverting an output of X decoder ADX applied on node 9, and an n-FET 12 responsive to an output of inverter circuit 16 for discharging the potential of word line 3 (WLi) to the level of ground potential via node 13.
n-FET 14 receives an internal operating power supply voltage Vcc at its gate. Word line driving signal .phi.x applied to node 10 is a signal boosted to a potential level higher than internal operating power supply voltage Vcc. Here, the potential of node 15 rises (due to capacitive coupling between the gate and drain of the n-FET) by the self-bootstrap function of n-FET 11. n-FET 14 is provided to prevent the boosted potential of node 15 from being transmitted to node 9. In other words, n-FET 14 functions as a decoupling transistor.
Inverter circuit 16 has a CMOS structure. The operating power supply voltage is set to internal operating power supply voltage Vcc (not shown) applied to node 8.
Word line driver WDi functions to receive a word line selecting signal of the level of internal operating power supply voltage Vcc from X decoder ADXi and apply the ability of driving word line 3 to this signal.
In order to generate a word line driving signal .phi.x of a boosted potential level, a high voltage generation circuit HVG and a word line driving signal generation circuit HSG are provided. High voltage generation circuit HVG responds to a pulse-like repetitive signal .phi.c to boost constantly internal operating power supply voltage Vcc for generating a boosted high voltage Vpp. Word line driving signal generation circuit HSG responds to a clock signal .phi.x0 for transmitting high voltage Vpp generated from high voltage generation circuit HVG to a node 18 as a word line driving signal .phi.x. A clock signal .phi.x0 is generated at a timing before the output potential of X decoder ADXi is ascertained (generated after a predetermined time elapse in response to the fall of row address strobe signal /RAS).
The repetitive signal .phi.c is generated from an on-chip ring oscillator or provided from an external source.
High voltage generation circuit HVG includes an n-FET 29 provided between internal power supply voltage node 8 and node 32, an n-FET 30 provided between nodes 35 and 27, a capacitor 31 provided between nodes 28 and 35, and a capacitor 33 provided between output node 27 and a second power supply voltage supplying source (ground potential source).
n-FET 29 has its gate and drain connected to charge the node 32 according to internal operating power supply voltage Vcc applied to node 8. n-FET 30 has the gate and drain connected and functions as a diode. Capacitor 31 provides capacitive coupling between nodes 28 and 35. Node 35 (node 32) is accompanied with a parasitic capacitance 34. Capacitor 33 serves to stabilize a high voltage Vpp generated at output node 27. Capacitor 31 serves to boost the potential level of node 35 by the repetitive signal .phi.c. High voltage generation circuit HVG generates high voltage Vpp of a voltage level higher than internal power supply voltage Vcc by the charge pumping function of capacitor 31.
Word line driving signal generation circuit HSG includes a p channel insulating gate type field effect transistor (referred to simply as "p-FET" hereinafter) 23 provided between nodes 17 and 25, a p-FET 20 provided between nodes 17 and 22, an n-FET 24 responsive to a control signal .phi.x0 provided to node 19 for discharging node 25 to the level of ground potential, an inverter circuit 26 for inverting control signal .phi.x0, and an n-FET 21 responsive to an output of inverter circuit 26 for reducing the potential of node 22 to ground level. p-FET 23 and p-FET 20 have their gates and drains cross-coupled. High voltage Vpp from high voltage generation circuit HVG is transmitted to node 17. A word line driving signal .phi.x is generated to an output node 18 of word line driving signal generation circuit HSG. Word line driving signal generation circuit HSG functions to convert control signal .phi.x0 of the level of internal operating power supply voltage Vcc applied to node 19 to a word line driving signal .phi.x of a level of high voltage Vpp. The structure of such a circuit with this function is disclosed in Japanese Patent Laying Open No. 49-114337.
High voltage generation circuit HVG and word line driving signal generation circuit HSG are included in control signal generating peripheral circuitry CSG shown in FIG. 25. Inverter circuit 26 has a CMOS structure to operate with internal operating power supply voltage Vcc as the operating power supply voltage. High voltage generation circuit HVG and word line driving signal generation circuit HSG are provided in common to respective word lined drivers provided for word line 3 (word lines WL0-WLn). The operation of each circuit shown in FIG. 27 will be described hereinafter.
First, the operation of high voltage generation circuit HVG will be described with reference to FIG. 28 which is an operation waveform diagram thereof. A repetitive signal .phi.c applied to node 28 is generated from an on-chip or external oscillation circuit utilizing ring oscillation. The signal .phi.c is assumed to be a pulse signal having a predetermined period and pulse width.
When internal operating power supply voltage Vcc is applied to internal operating power supply voltage terminal 8, the potentials of nodes 32 and 35 are charged to the potential level of Vcc -VTN by charging n-FET 29. Here, VTN is the threshold voltage of n-FET 29. By rectifying n-FET 30, the potential level of node 27 is charged to the potential level of Vcc- 2.multidot.VTN.
The provision of repetitive signal .phi.c to node 28 causes initiation of a boosting operation in high voltage generation circuit HVG. For the sake of simplification, it is assumed that the boosting operation in high voltage generation circuit HVG is initiated after the potential level of node 32 and output node 27 are stabilized to the above-described potential levels of Vcc-VTN and Vcc-2.multidot.VTN.
When repetitive signal .phi.c rises after the potentials of node 32 and output node 27 attain the level of Vcc-VTN and Vcc-2.multidot.VTN, respectively, charge is introduced to node 35 via boosting capacitor 31. Therefore, the potential of node 35 rises. This potential rise of node 35 causes charge to be provided to output node 27 via n-FET 30, whereby potential V27 of output node 27 is raised by: EQU .DELTA.V27=C31.multidot.Vcc/(C31+C33)
Here, C31 represents the capacitance value of boosting capacitor 31, and C33 represents the capacitance value of stabilizing capacitor 33.
When repetitive signal .phi.c falls, the potential of node 32 (node 35) is decreased due to capacitive coupling by boosting capacitor 31. However, because n-FET 30 has the gate and drain connected to serve as a diode, it enters a non-conducting state, and potential V27 of output node 27 is not decreased and maintains the potential boosted at the time of the prior rise of repetitive signal .phi.c. In response to the fall of repetitive signal .phi.c, the potentials of nodes 32 and 35 are charged by n-FET 29 to be restored to the level of Vcc-VTN.
By repeating the above-described operation, charges are injected into nodes 32 and 35 via boosting capacitor 31. Charge is introduced to output node 27 via n-FET 30 every time the potential thereat rises. As a result, the potential of output node 27 gradually rises.
The eventual potential V32max of node 32 (node 35) is as follows: EQU V32max=(Vcc-VTN)+C31.multidot.Vcc/(C31+C34)
where C34 represents the capacitance value of parasitic capacitance 34. Here, potential V27 of output node 27 attains a value lower by the threshold voltage VTN of n-FET 30 than the potential V32 (=V35) of node 32 (node 35). More specifically, the eventual potential V27max of output node 27 is as follows: EQU V27max=V32max-VTN EQU =(Vcc-2.multidot.VTN)+C31.multidot.Vcc/(C31+C34)
In practice, capacitance value C31 of boosting capacitor 31 can easily be made high enough in comparison with capacitance value C34 of parasitic capacitance 34. Therefore, the third term in the above two equations are approximately equal to internal operating power supply voltage Vcc. Suppose Vcc=3.3 V, and VTN=0.8 V, the potential V27max of output node 27 is as set forth in the following from the above equation. EQU V27max=2(Vcc-VTN)=5.0(V)
More specifically, potential V27max of output node 27 is approximately 1.5 times the internal operating power supply voltage Vcc. This high voltage is stabilized by a stabilizing capacitor 33 having a great capacitance value.
The operation of a word line driving signal generation circuit and a word line driver will be described with reference to FIG. 29 which is a waveform diagram thereof.
At time t0 when control signal .phi.x0 attains a L level (logical low), n-FET 24 is turned off and n-FET 21 is turned on by inverter circuit 26. Therefore, the potential of node 22 attains the L level of ground potential, and the potential of node 25 attains the level of high voltage Vpp applied to node 17 via p-FET 23. When the potential of node 25 attains the level of high voltage Vpp, p-FET 20 is completely turned off, so that the potential of node 22 is reliably discharged to the level PG,16 of ground potential via n-FET 20. The potential of word line driving signal .phi.x reliably attains the level of ground potential.
In word line driver WDi, the output potential (the potential of node 9) of X decoder ADXi attains a L level (ground potential level), whereby n-FET 12 and n-FET 11 are turned on and off, respectively. As a result, the potential level of word line driving signal .phi.xi on word line 3 attains the L level of ground potential.
Next, when row address strobe signal /RAS (refer to FIG. 25) falls to a L level, a row selecting operation is initiated. In response to the fall of row address strobe signal /RAS, X decoder ADX (refer to FIG. 25) executes a row selecting operation. Here, it is assumed that (unit) X decoder ADXi shown in FIG. 27 is selected.
At time t1 when the potential level of node 9 rises to the level of internal power supply voltage Vcc, the output of inverter circuit 16 of word line driver WDi attains a L level of ground potential, whereby n-FET 12 is turned off from an on state. Also, node 15 is charged from node 9 via n-FET 14 to increase in potential. Between nodes 9 and 15, n-FET 14 is provided having the gate connected to power supply voltage node 8 supplying internal operating power supply voltage Vcc. Therefore, the potential level of node 15 rises to the potential level of Vcc-VTN. Here, VTN is the threshold voltage of n-FET 14. Therefore n-FET 11 is turned on, and word line 3 is discharged via n-FET 11 and n-FET 21 to maintain the level of ground potential.
When the potential level of node 9 is stabilized, control signal .phi.x0 applied to node 19 rises to the H level at time t2. In response to the fall of row address strobe signal /RAS control signal .phi.x0 rises to the level of internal operating power supply voltage Vcc after a predetermined time. The rise of control signal .phi.x0 to the level of internal operating power supply voltage Vcc causes n-FET 24 to be turned on and n-FET 21 to be turned off. As a result, node 25 is discharged to the level of ground potential by n-FET 24. In response, p-FET 20 is turned on to raise the potential of node 22. Eventually, when p-FET 23 is turned off and node 25 falls to the level of ground potential, the potential level of node 22 attains the level of high voltage Vpp applied to node 17 via p-FET 20. As a result, a word line driving signal .phi.x is generated.
When word line driving signal .phi.x of the level of high voltage Vpp is applied to node 10 in word line driver WDi, the potential level of node 15 rises by the voltage change of node 10 (by high voltage Vpp) due to the self bootstrap function of n-FET 11 (by the capacitive coupling between the gate and drain of n-FET). As a result, the potential level of node 15 becomes Vcc-VTN+Vpp, i.e. exceeds Vpp+VTN. Therefore, there is no loss of the threshold voltage in n-FET 11, and the potential level of word line driving signal .phi.xi transmitted on word line 3 rises to the level of high voltage Vpp.
The rise of word line driving signal .phi.xi transmitted to word line 3 up to the level of high voltage Vpp causes the selecting transistor 5 in memory cell 1 to attain a deep on state rapidly. Therefore, the charge stored in memory capacitor 6 is transmitted on bit line 2 without loss of the threshold voltage of selecting transistor (n-FET) 5.
Then, the sensing operation of the sense amplifier is carried out, followed by reading or writing of data of a selected memory cell.
When one memory cycle is completed, control signal .phi.x0 falls to the L level and the output of X decoder ADXi also falls to the L level at time t3. The potentials of each signal and node are restored to the state of time t0.
The charging operation of word line 3, i.e. potential increase thereof, will be described in detail hereinafter.
The charging operation of word line 3 is realized by the transfer of charge from stabilizing capacitor 33 in high voltage generation circuit HVG to parasitic capacitance 7 of word line 3. Therefore, the potential level of output node 27 of high voltage generation circuit HVG is slightly reduced since charge is transferred to word line 3 at the time of word line selection. However, if the capacitance value of stabilizing capacitor 33 is set to a value sufficiently greater than the capacitance value of parasitic capacitance 7 of word line 3, there is almost no decrease in the potential level of output node 27. Therefore, the potential level of a selected word line can maintain the level of high voltage Vpp.
Because the potential level V(WL) of word line 3 can be obtained by: EQU V(WL)=C33.multidot.Vpp/(C33+C7)
the potential level on word line 3 can take the level of high voltage Vpp if the capacitance value C7 of parasitic capacitance 7 is small enough to be neglected in comparison with the capacitance value C33 of stabilizing capacitor 33.
From the standpoint of high density and high integration density, stabilizing capacitor 33 must have superior space efficiency that can realize a relatively high capacitance value with a small occupying area. AMOS capacitor utilizing an insulating gate type field effect transistor is generally employed as such capacitors.
FIG. 30 is a MOS capacitor wherein (A) shows a sectional view thereof, (B) shows an electrical connection circuit thereof, and (C) shows an electrical equivalent circuit thereof.
Referring to FIG. 30 (A), a MOS capacitor includes N type impurity regions 102a and 102b formed at a predetermined region on a P type semiconductor substrate 101, a gate insulating film (capacitor insulating film) 104 formed on the surface of semiconductor Substrate 101, and a gate electrode 103 formed on gate insulating film 104. Impurity regions 102a and 102b provide one electrode leading portion of the capacitor (in FIG. 30 (A), the electrode leading portion connected to ground potential GND, i.e. ground line). Gate electrode 103 implements the other electrode of the capacitor, and is formed of polycrystalline silicon, refractory metal silicide such as molybdenum silicide or tungsten silicide, or a multilayer structure of polycrystalline silicon and refractory metal silicide.
Gate electrode 103 is connected to output node 27 receiving high voltage Vpp. The power supply and ground lines between gate electrode 103 and output node 27 are formed of a low resistance metal such as aluminum. Gate insulating film 104 is formed using an insulating film such as silicon dioxide (SiO.sub.2) film. The source and drain electrode 108 is formed of a low resistance conductor such as aluminum to provide electrical contact with impurity regions 102a and 102b. Ground potential GND from the ground line is applied to impurity regions 102a and 102b.
An interlayer insulating film 109 is provided to electrically insulate the electrodes 103 and 108 from each other.
When a voltage exceeding the threshold voltage is applied to gate electrode 103, an inversion layer (N type inversion layer) 101 is formed at the surface of semiconductor substrate 101. This inversion layer 101 forms one electrode of the capacitor. More specifically, in the MOS capacitor shown in FIG. 30(A), one electrode of the capacitor is inversion layer 101, and the other electrode is gate electrode 103. Ground potential GND is applied to inversion layer 101 via impurity region 102. When connection of one electrode to ground potential GND is realized and high voltage Vpp is applied to the other electrode (gate electrode 103), this capacitor functions as the stabilizing capacitor shown in FIG. 27.
A MOS capacitor has a structure identical to a MOS transistor (insulating gate type field effect transistor) used within a memory chip. It can be regarded as a MOS transistor having the source electrode and the drain electrode connected in common to ground potential GND (refer to FIG. 30(B) and (C)).
A capacitor of a MOS structure shown in FIG. 30(A) is used because the thickness of a dielectric film (capacitor insulating film) can be reduced. It also has the advantage of forming a capacitor with superior space efficiency by using an empty region of a nearby circuit.
In a general dynamic random access memory, the potential of a selected word line is required to be a level of approximately 1.5 times the internal operating power supply voltage (to be more precise, the potential level corresponding to the data of the high level written in the memory cell) in order to increase the speed of reading data from a memory cell and writing data of a sufficient potential level to a memory cell.
The FET (insulating gate type field effect transistor) used in a dynamic random access memory has the thickness of the gate insulating film determined depending upon performance of its operation speed and stability. For example, when the operating power supply voltage is 3.3 V, the gate insulating film of a FET is set to approximately 100 .ANG..
In this case, the electric field E applied to the gate insulating film is: EQU E=V/t=3.3.multidot.1.5/100=5.multidot.10.sup.6 [V/cm]
An electric field sufficiently lower than the dielectric breakdown voltage of 10.multidot.10.sup.6 V/cm is applied to the gate insulating film. Thus, reliability of a gate insulating film is ensured.
An operation lifetime test was carried out on such dynamic random access memories. It was found out that the life time of a dynamic random access memory with an operating power supply voltage of 3.3 V is shorter than that of a dynamic random access memory with an operating power supply voltage of 5 V. The operation life time test for a 3.3 V dynamic random access memory was carried out under the conditions of an ambient temperature of 125.degree. C., and internal operating power supply voltage Vcc of 5 V (Vpp =5.times.1.5=7.5 (V)), and an operating time period of 1000 hours. The operation life time test for a 5 V dynamic random access memory was carried out under the conditions of an ambient temperature of 125.degree. C., an internal operating power supply voltage Vcc of 7.5 V (Vpp=7.5.times.1.5=11.25 (V)).
Then, the operating conditions were set to a more severe level. An accelerated operation life test was carried out in which an operation life time test of 1000 hours corresponds to that of 1500 hours. Although there was little difference in the defective rate of DRAMs having Vcc of 3.3 V and 5.0 V in an operation life test of a general standard, there were cases where the defective rate of a dynamic random access memory with an operating power supply voltage of 3.3 V was higher than that of a dynamic random access memory with an operating power supply voltage of 5 V in the accelerated operation life time test.
This cause was traced to the fact that dielectric breakdown occurred in the capacitor used for high voltage stabilization, resulting in short circuit between the high voltage output node and ground potential, to prevent a high voltage from being generated in stability. Such dielectric breakdown is considered to be caused by an electric field greater than the tolerable value being applied due to variation in the thickness of the gate insulating film in the manufacturing process. Even if the reference value of variation in the film thickness is 5 .ANG., a thinner film will suffer a greater effect.
In a dynamic random access memory, the size of a FET is reduced for the sake of high density, high integration density, and low cost even if the operating power supply voltage is the same 3.3 V. Accordingly, the thickness of the gate insulating film is further reduced to 90 .ANG., 80 .ANG., etc. Therefore, the insulation characteristics of a capacitor for high voltage stabilization must be improved.
Even if instantaneous dielectric breakdown due to application of a high electric field does not occur, it is known that fatigue failure of an insulating film (time dependent dielectric breakdown: TDDB) due to stress of an electric field application over a long time will occur. There is a possibility of dielectric breakdown even when an electric field that does not induce dielectric breakdown is applied. Thus, reliability of the insulating characteristics of a capacitor for high voltage stabilization must be ensured.
Such problem of insulating characteristics not only in a stabilizing capacitor, but also in a MOS transistor, is not limited to a semiconductor memory device, and is also seen in integrated circuit devices utilizing high voltage.